RedisOrm supposed to be almost drop-in replacement of ActiveRecord 2.x. It's based on the Redis - advanced key-value store and is work in progress.

Here's the standard model definition:

class User < RedisOrm::Base
  property :first_name, String
  property :last_name, String

  timestamps

  # OR
  # property :created_at, Time
  # property :modified_at, Time

  index :last_name
  index [:first_name, :last_name]

  has_many :photos
  has_one :profile

  after_create :create_mailboxes

  def create_mailboxes
    # ...
  end
end

Installing redis_orm

stable release:

gem install redis_orm

or edge version:

git clone git://github.com/german/redis_orm.git
cd redis_orm
bundle install

To run the tests you should have redis installed already. Please check Redis download/installation page.

rspec

Setting up a connection to the redis server

If you are using Rails you should initialize redis and set up global $redis variable in config/initializers/redis.rb file:

require 'redis'
$redis = Redis.new(:host => 'localhost', :port => 6379)

Defining a model and specifing properties

To specify properties for your model you should use the following syntax:

class User < RedisOrm::Base
  property :first_name, String
  property :last_name, String
  property :created_at, Time
  property :modified_at, Time
end

Supported property types:

  • Integer

  • String

  • Float

  • RedisOrm::Boolean there is no Boolean class in Ruby so it's a special class to store TrueClass or FalseClass objects

  • Time or DateTime

  • Array or Hash RedisOrm automatically will handle serializing/deserializing arrays and hashes into strings using Marshal class

Following options are available in property declaration:

  • :default

    The default value of the attribute when it's getting saved w/o any.

  • :sortable

    if true is specified then you could sort records by this property later

Note that when you're using :sortable option redis_orm maintains one additional list per attribute. Also note that the #create method could be 3 times slower in some cases (this will be improved in future releases), while the #find performance is basically the same (see the "benchmarks/sortable_benchmark.rb").

Expiring record after certain period of time

You could expire record stored in Redis by specifying TTL in seconds invoking expire method of the class like this:

class PhantomUser < RedisOrm::Base
  property :name, String
  property :persist, RedisOrm::Boolean, :default => true

  expire 15.minutes.from_now
end

Also you could specify a condition when expire would be set on record's key:

  expire 15.minutes.from_now, :if => Proc.new {|r| !r.persist?}

Also you could override class method expire by using expire_in key when saving object:

ExpireUser.create :name => "Ghost record", :expire_in => 50.minutes.from_now

Searching records by the value

Usually it's done via declaring an index and using :conditions hash or dynamic finders. For example:

class User < RedisOrm::Base
  property :name, String

  index :name
end

User.create :name => "germaninthetown"

# via dynamic finders:
User.find_by_name "germaninthetown" # => found 1 record
User.find_all_by_name "germaninthetown" # => array with 1 record

# via *:conditions* hash:
User.find(:all, :conditions => {:name => "germaninthetown"}) # => array with 1 record
User.all(:conditions => {:name => "germaninthetown"}) # => array with 1 record

Dynamic finders work mostly the way they work in ActiveRecord. The only difference is if you didn't specified index or compound index on the attributes you are searching on the exception will be raised. So you should make an initial analysis of model and determine properties that should be searchable.

Options for #find/#all

To extract all or part of the associated records you could use 4 options:

  • :limit

  • :offset

  • :order

Either :desc or :asc (default), since records are stored with Time.now.to_f scores, by default they could be fetched only in that (or reversed) order. To order by different property you should:

  1. specify :sortable => true as option in property declaration

  2. specify the property by which you wish to order :order => [:name, :desc] or :order => [:name] (:asc order is default)

  • :conditions

Hash where keys must be equal to the existing property name (there must be index for this property too).

# for example we associate 2 photos with the album
@album.photos << Photo.create(:image_type => "image/png", :image => "boobs.png")
@album.photos << Photo.create(:image_type => "image/jpeg", :image => "facepalm.jpg")

@album.photos.all(:limit => 0, :offset => 0) # => []
@album.photos.all(:limit => 1, :offset => 0).size # => 1
@album.photos.all(:limit => 2, :offset => 0) # [...]
@album.photos.all(:limit => 1, :offset => 1, :conditions => {:image_type => "image/png"})
@album.photos.find(:all, :order => "asc")

Photo.find(:first, :order => "desc")
Photo.all(:order => "asc", :limit => 5)
Photo.all(:order => "desc", :limit => 10, :offset => 50)
Photo.all(:order => "desc", :offset => 10, :conditions => {:image_type => "image/jpeg"})

Photo.find(:all, :conditions => {:image => "facepalm.jpg"}) # => [...]
Photo.find(:first, :conditions => {:image => "boobs.png"}) # => [...]

Using UUID instead of numeric id

You could use universally unique identifiers (UUIDs) instead of a monotone increasing sequence of numbers as id/primary key for your models.

Example of UUID: b57525b09a69012e8fbe001d61192f09.

To enable UUIDs you should invoke use_uuid_as_id class method:

class User < RedisOrm::Base
  use_uuid_as_id

  property :name, String

  property :created_at, Time
end

UUID gem is installed as a dependency.

An excerpt from https://github.com/assaf/uuid :

UUID (universally unique identifier) are guaranteed to be unique across time and space.

A UUID is 128 bit long, and consists of a 60-bit time value, a 16-bit sequence number and a 48-bit node identifier.

Note: when using a forking server (Unicorn, Resque, Pipemaster, etc) you don’t want your forked processes using the same sequence number. Make sure to increment the sequence number each time a worker forks.

For example, in config/unicorn.rb:

after_fork do |server, worker|
  UUID.generator.next_sequence
end

Indices

Indices are used in a different way then they are used in relational databases. In redis_orm they are used to find record by they value rather then to quick access them.

You could add index to any attribute of the model (index also could be compound):

class User < RedisOrm::Base
  property :first_name, String
  property :last_name, String

  index :first_name
  index [:first_name, :last_name]
end

With index defined for the property (or properties) the id of the saved object is stored in the sorted set with special name, so it could be found later by the value. For example with defined User model from the above code:

user = User.new :first_name => "Robert", :last_name => "Pirsig"
user.save

# 2 redis keys are created "user:first_name:Robert" and "user:first_name:Robert:last_name:Pirsig" so we could search records like this:

User.find_by_first_name("Robert")                             # => user
User.find_all_by_first_name("Robert")                         # => [user]
User.find_by_first_name_and_last_name("Robert", "Pirsig")     # => user
User.find_all_by_first_name_and_last_name("Chris", "Pirsig")  # => []

Indices on associations are also created/deleted/updated when objects with has_many/belongs_to associations are created/deleted/updated (excerpt from association_indices_test.rb):

class Article < RedisOrm::Base
  property :title, String
  has_many :comments
end

class Comment < RedisOrm::Base
  property :body, String
  property :moderated, RedisOrm::Boolean, :default => false
  index :moderated
  belongs_to :article
end

article = Article.create :title => "DHH drops OpenID on 37signals"
comment1 = Comment.create :body => "test"    
comment2 = Comment.create :body => "test #2", :moderated => true

article.comments << [comment1, comment2]

# here besides usual indices for each comment, 2 association indices are created so #find with *:conditions* on comments should work

article.comments.find(:all, :conditions => {:moderated => true})
article.comments.find(:all, :conditions => {:moderated => false})

Index definition supports following options:

  • :unique Boolean default: false

If true is specified then value is stored in ordinary key-value structure with index as the key, otherwise the values are added to sorted set with index as the key and Time.now.to_f as a score.

  • :case_insensitive Boolean default: false

If true is specified then property values are saved downcased (and then are transformed to downcase form when searching). Works for compound indices too.

Associations

RedisOrm provides 3 association types:

  • has_one

  • has_many

  • belongs_to

HABTM association could be emulated with 2 has_many declarations in related models.

has_many/belongs_to associations

class Article < RedisOrm::Base
  property :title, String
  has_many :comments
end

class Comment < RedisOrm::Base
  property :body, String
  belongs_to :article
end

article = Article.create :title => "DHH drops OpenID support on 37signals"
comment1 = Comment.create :body => "test"
comment2 = Comment.create :body => "test #2"

article.comments << [comment1, comment2]

# or rewrite associations
article.comments = [comment1, comment2]

article.comments # => [comment1, comment2]
comment1.article # => article
comment2.article # => article

Backlinks are automatically created.

has_one/belongs_to associations

class User < RedisOrm::Base
  property :name, String
  has_one :profile  
end

class Profile < RedisOrm::Base
  property :age, Integer

  validates_presence_of :age  
  belongs_to :user
end

user = User.create :name => "Haruki Murakami"
profile = Profile.create :age => 26
user.profile = profile

user.profile # => profile
profile.user # => user

Backlink is automatically created.

has_many/has_many associations (HABTM)

class Article < RedisOrm::Base
  property :title, String
  has_many :categories
end

class Category < RedisOrm::Base
  property :name, String
  has_many :articles
end

article = Article.create :title => "DHH drops OpenID support on 37signals"

cat1 = Category.create :name => "Nature"
cat2 = Category.create :name => "Art"
cat3 = Category.create :name => "Web"

article.categories << [cat1, cat2, cat3]

article.categories # => [cat1, cat2, cat3]
cat1.articles # => [article]
cat2.articles # => [article]
cat3.articles # => [article]

Backlinks are automatically created.

Self-referencing association

class User < RedisOrm::Base
  property :name, String
  index :name
  has_many :users, :as => :friends
end

me = User.create :name => "german"
friend1 = User.create :name => "friend1"
friend2 = User.create :name => "friend2"

me.friends << [friend1, friend2]

me.friends # => [friend1, friend2]
friend1.friends # => []
friend2.friends # => []

As an exception if :as option for the association is provided the backlinks aren't created.

Polymorphic associations

Polymorphic associations work the same way they do in ActiveRecord (2 keys are created to store type and id of the record)

class CatalogItem < RedisOrm::Base
  property :title, String

  belongs_to :resource, :polymorphic => true
end

class Book < RedisOrm::Base
  property :price, Integer
  property :title, String

  has_one :catalog_item
end

class Giftcard < RedisOrm::Base
  property :price, Integer
  property :title, String

  has_one :catalog_item
end

book = Book.create :title => "Permutation City", :author => "Egan Greg", :price => 1529
giftcard = Giftcard.create :title => "Happy New Year!"

ci1 = CatalogItem.create :title => giftcard.title
ci1.resource = giftcard

ci2 = CatalogItem.create :title => book.title
ci2.resource = book

All associations supports following options:

  • :as

Symbol Association could be accessed by provided name

  • :dependent

Symbol could be either :destroy or :nullify (default value)

Clearing/reseting associations

You could clear/reset associations by assigning appropriately nil/[] to it:

# has_many association
@article.comments << [@comment1, @comment2]
@article.comments.count # => 2
@comment1.article       # => @article

# clear    
@article.comments = []
@article.comments.count # => 0
@comment1.article       # => nil

# belongs_to (same for has_one)
@article.comments << [@comment1, @comment2]
@article.comments.count # => 2
@comment1.article       # => @article

# clear
@comment1.article = nil
@article.comments.count # => 1
@comment1.article       # => nil

For more examples please check test/associations_test.rb and test/polymorphic_test.rb

Validation

RedisOrm includes ActiveModel::Validations. So all well-known validation callbacks are already in. An excerpt from test/validations_test.rb:

class Photo < RedisOrm::Base
  property :image, String

  validates_presence_of :image
  validates_length_of :image, :in => 7..32
  validates_format_of :image, :with => /\w*\.(gif|jpe?g|png)/
end

Callbacks

RedisOrm provides 6 standard callbacks:

after_save :callback
before_save :callback
after_create :callback
before_create :callback
after_destroy :callback
before_destroy :callback

They are implemented differently than in ActiveModel though work as expected:

class Comment < RedisOrm::Base
  property :text, String

  belongs_to :user

  before_save :trim_whitespaces

  def trim_whitespaces
    self.text = self.text.strip
  end
end

Saving records

When saving object standard ActiveModel's #valid? method is invoked at first. Then appropriate callbacks are run. Then new Hash in Redis is created with keys/values equal to the properties/values of the saving object.

The object's id is stored in "model_name:ids" sorted set with Time.now.to_f as a score. So records are ordered by created_at time by default. Then record's indices are created/updated.

Dirty

Redis_orm also provides dirty methods to check whether the property has changed and what are these changes. To check it you could use 2 methods: #property_changed? (returns true or false) and #property_changes (returns array with changed values).

File attachment management with paperclip and redis

3 simple steps you should follow to manage your file attachments with redis and paperclip.

Tests

Though I'm a big fan of the Test::Unit all tests are based on RSpec. And the only reason I use RSpec is possibility to define before(:all) and after(:all) hooks. So I could spawn/kill redis-server's process (from test_helper.rb):

RSpec.configure do |config|
  config.before(:all) do
    path_to_conf = File.dirname(File.expand_path(__FILE__)) + "/redis.conf"
    $redis_pid = spawn 'redis-server ' + path_to_conf, :out => "/dev/null"
    sleep(0.3) # must be some delay otherwise "Connection refused - Unable to connect to Redis"
    path_to_socket = File.dirname(File.expand_path(__FILE__)) + "/../redis.sock"
    $redis = Redis.new(:host => 'localhost', :path => path_to_socket)
  end

  config.before(:each) do
    $redis.flushall if $redis
  end

  config.after(:each) do
   $redis.flushall if $redis
  end

  config.after(:all) do
    Process.kill 9, $redis_pid.to_i if $redis_pid
  end
end

To run all tests just invoke rake test

Contributors

Tatsuya Sato

Copyright © 2011 Dmitrii Samoilov, released under the MIT license

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.